weka.classifiers.bayes.blr

Class LaplacePriorImpl

java.lang.Object

weka.classifiers.bayes.blr.Prior

weka.classifiers.bayes.blr.LaplacePriorImpl

All Implemented Interfaces:

java.io.Serializable, RevisionHandler

public class LaplacePriorImpl
extends Prior

Implementation of the Gaussian Prior update function based on modified CLG Algorithm (CLG-Lasso) with a certain Trust Region Update based on Laplace Priors.

Version:

$Revision: 1.2 $

Author:

Navendu Garg(gargnav@iit.edu)

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
LaplacePriorImpl()

Method Summary

Modifier and Type	Method and Description
void	computeLogLikelihood(double[] betas, Instances instances) Computes the log-likelihood values using the implementation in the Prior class.
void	computePenalty(double[] betas, double[] hyperparameters) This function computes the penalty term specific to Laplacian distribution.
java.lang.String	getRevision() Returns the revision string.
double	laplaceUpdate(int j, double sign) This is the CLG-lasso update function described in the
double	update(int j, Instances instances, double beta, double hyperparameter, double[] r, double deltaV) Update function specific to Laplace Prior.

Methods inherited from class weka.classifiers.bayes.blr.Prior

computelogLikelihood, getLoglikelihood, getLogPosterior, getPenalty

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

LaplacePriorImpl

public LaplacePriorImpl()

Method Detail

update

public double update(int j,
                     Instances instances,
                     double beta,
                     double hyperparameter,
                     double[] r,
                     double deltaV)

Update function specific to Laplace Prior.

Overrides:

update in class Prior

laplaceUpdate

public double laplaceUpdate(int j,
                            double sign)

This is the CLG-lasso update function described in the

 @TechReport{blrtext04,
author = {Alexander Genkin and David D. Lewis and David Madigan},
title = {Large-scale bayesian logistic regression for text categorization},
institution = {DIMACS},
year = {2004},
url = "http://www.stat.rutgers.edu/~madigan/PAPERS/shortFat-v3a.pdf",
OPTannote = {}
}

Parameters:

j -

Returns:

double value

computeLogLikelihood

public void computeLogLikelihood(double[] betas,
                                 Instances instances)

Computes the log-likelihood values using the implementation in the Prior class.

Parameters:

betas -

instances -

hyperparameter -

computePenalty

public void computePenalty(double[] betas,
                           double[] hyperparameters)

This function computes the penalty term specific to Laplacian distribution.

Overrides:

computePenalty in class Prior

Parameters:

betas -

hyperparameters -

getRevision

public java.lang.String getRevision()

Returns the revision string.

Returns:

the revision